What are the four metric types in Prometheus and how to choose between them?

Four metric types in Prometheus:

1. Counter:

   • Cumulative value that only increases
   • Used for recording request counts, error counts
   • Example: http_requests_total
   • Common functions: rate() for calculating growth rate

2. Gauge:

   • Instantaneous value that can increase or decrease
   • Used for recording temperature, memory usage
   • Example: memory_usage_bytes
   • Common functions: avg(), max(), min()

3. Histogram:

   • Samples observations and groups them into buckets
   • Provides _count, _sum, _bucket metrics
   • Used for recording request latency, response size
   • Example: http_request_duration_seconds
   • Can calculate quantiles

4. Summary:

   • Similar to Histogram but calculates quantiles on the client side
   • Provides _count, _sum, and quantile metrics
   • Suitable for scenarios with smaller data volumes
   • Example: rpc_duration_seconds

Selection Guidelines:

• Use Histogram for aggregated statistics
• Use Summary for precise quantiles
• Use Counter for cumulative values
• Use Gauge for instantaneous values